Long-Term Outcome of Iatrogenic Gas Embolism Nicolas Genotelle Cendrine Chabbaut Anne Huon Alexis Tabah Je´Roˆme Aboab Sylvie Chevret Djillali Annane

Intensive Care Med DOI 10.1007/s00134-010-1821-9 ORIGINAL

Jacques Bessereau Long-term outcome of iatrogenic gas embolism Nicolas Genotelle Cendrine Chabbaut Anne Huon Alexis Tabah Je´roˆme Aboab Sylvie Chevret Djillali Annane

Received: 19 June 2009 Abstract Objective: To establish of 1-year mortality (OR = 4.39, 95% Accepted: 24 January 2010 the incidence and long-term progno- CI 1.46–12.20 and OR = 6.30, 1.71– sis of iatrogenic gas embolism. 23.21, respectively). Among ICU Ó Copyright jointly held by Springer and Methods: This was a prospective survivors, independent predictors of ESICM 2010 inception cohort. We included all 1-year mortality were age consecutive adults with proven iatro- (OR = 1.07, 1.01–1.14), Babinski genic gas embolism admitted to the sign (OR = 6.58, 1.14–38.20) and sole referral academic hyperbaric acute kidney failure (OR = 8.09, center in Paris. Treatment was stan- 1.28–51.21). Focal motor deﬁcits dardized as one hyperbaric session at (OR = 12.78, 3.98–41.09) and 4 ATA for 15 min followed by two Babinski sign (OR = 6.76, 2.24– 45-min plateaus at 2.5 then 2 ATA. 20.33) on ICU admission, and dura- J. Bessereau Á N. Genotelle Á A. Huon Á Inspired fraction of oxygen was set at tion of mechanical ventilation of A. Tabah Á J. Aboab Á D. Annane ()) 100% during the entire dive. Primary 5 days or more (OR = 15.14, General Intensive Care Unit, Hyperbaric endpoint was 1-year mortality. All 2.92–78.52) were independent pre- Centre, Raymond Poincare´ Hospital patients had evaluation by a neurolo- dictors of long-term sequels. (AP-HP), University of Versailles SQY Conclusions: Gas embolism (UniverSud Paris), 104 boulevard Raymond gist, visual ﬁeld tested by Goldman Poincare´, 92380 Garches, France kinetic perimetry and brain MRI or complicates 2.65 per 100,000 hospi- e-mail: [email protected] CT scan at 6 months and 1 year. talizations, and is associated with Tel.: ?33-1-47107786 Results: From January 1993 to high mortality and morbidity. Babin- Fax: ?33-1-47107783 August 2004, 125 of 4,727,496 hos- ski sign on ICU admission is pitalizations had proven iatrogenic associated with poor prognosis. C. Chabbaut Á S. Chevret gas embolism. The crude mortality De´partement de Biostatistiques et Keywords Iatrogenic Á Informatique Me´dicale, AP-HP, was 25/119 (21%) at 1 year. Cardiac Erm321-INSERM, Saint-Louis Hospital arrest at time of accident and ICU Gas embolism Á Outcome Á Mortality Á (AP-HP), 10 avenue Vellefaux, admission, and SAPS II of 33 or more Sequels Á Glasgow Outcome Scale 75010 Paris, France were independent prognostic factors

Introduction obstruction with subsequent ischemia/reperfusion inju- ries. This accident is commonly recognized when sudden Gas embolism is a rare complication of invasive medical neurological, pulmonary or cardiac abnormalities occur or surgical procedures [1]. Gas embolism may originate in during invasive procedures. More subtle cases may be the venous system or in the systemic arterial circulation. diagnosed by echocardiography, transcranial Doppler, CT In the former case, the gaseous embolus may reach the scan or other diagnostic procedures. systemic circulation via pulmonary shunting or via patent Gas embolism is considered as an emergency, and foramen ovale. When the gaseous embolus is in the treatment includes interruption of the embolic procedure, arterial circulation, it may cause transient arterial attempt to extract intravascular gas, external cardiac massage and immediate oxygenation, at best under coma, focal motor deficit, seizures, cardiac arrest, car- hyperbaric conditions. Hyperbaric oxygen therapy (HBO) diovascular collapse and acute dyspnea; (3) evidence of is recommended, although randomized, controlled clinical gas bubble entry in the systemic circulation as noticed trials are lacking [1–3]. Indeed, HBO may reduce gas during a surgical or radiological procedure, or presence of bubble volume by increasing ambient pressure and by gas bubbles in left heart cavities at echocardiography or in inducing blood denitrogenation, which facilitates oxygen cerebral arteries at transcranial Doppler or CT scan. diffusion from the emboli to the surrounding plasma. It Notable exclusion criteria included underlying disease also permits a better oxygenation of ischemic tissues by with a life expectancy of less than 1 month, moribund increasing dissolved-oxygen content in blood and tissues. patients likely to die within 24 h and patients with Finally, HBO may decrease cerebral edema. Potential decompression sickness. disadvantages of HBO include risks associated with the transport of critically ill patients, hyperoxic seizures, Hyperbaric oxygen therapy claustrophobic reactions and barotraumas. Little is known about the morbidity and mortality An identical HBO procedure was used for all patients. We associated with iatrogenic gas embolism. Indeed, the used a multiplace hyperbaric chamber (CxPRO, CO- diagnosis of gas embolism may be difficult in conditions MEXÒ, Marseille, France) pressurized with compressed other than high risk surgery. In addition, when recognized, air at 4 atmospheres absolute (ATA) for 15 min (30 msw iatrogenic gas embolism requires a specific management depth, 405 kPa) followed by two 45-min plateaus at 2.5 and patient referral to a hyperbaric facility. The long-term then 2 ATA (Appendix 1). All intubated patients were follow-up of patients often does not involve the hyperbaric mechanically ventilated in the chamber with 100% oxy- physicians. Thus, unsurprisingly, the few available retro- gen (RCH LAMAÒ). The remaining patients received spective studies showed a wide range of mortality from 5 100% oxygen during the full session via a tight-fitting to 23% and of sequels from 13 to 71% [4–14]. face mask. All patients received intravenous diazepam or We took the opportunity, as a referral center in Paris midazolam, and phenobarbital to prevent hyperoxic sei- and its suburb, to conduct a prospective cohort study with zures. The hyperbaric session was completed by 12 h of careful short-term and 1-year follow-up to determine the normobaric oxygen therapy. mortality and morbidity from proven iatrogenic gas embolism. Data collection Data collection at inclusion Materials and methods Clinical evaluation The following data were recorded: (1) This was a prospective single-center, inception cohort general characteristics including demographics, diagnosis study performed at Raymond Poincare´ University-affili- and recent surgery; (2) severity of illness assessed by vital ated Hospital. The intensive care unit (ICU) has a separate signs, Glasgow Coma Scale score and Simplified Acute unit for HBO. This is the sole public hyperbaric oxygen Physiology Score II (SAPS II) [15]; (3) neurological, facility for Paris and its suburb. Therefore, all cases of cardiac and respiratory symptoms; (4) type and source of iatrogenic gas embolism in this area are referred to our gas embolism; (5) time to HBO treatment; (6) presence center. The protocol was approved by the Comite´ Con- of a patent foramen ovale (PFO); (7) interventions sultatif de Protection des Personnes dans la Recherche including type and doses of vasopressors, and mechanical Biome´dicale of Saint-Germain en Laye (Yvelines, ventilation. France), and informed consent was waived according to the observational nature of the study; all patients received routine standard of care for gas embolism (cardiopulmo- Laboratory variables nary resuscitation, fluid therapy, mechanical ventilation and positive inotropic support whenever needed). Hematological and chemistry data, blood gas determina- tions, electroencephalography and whenever possible magnetic resonance imagings (MRI) or computerized Patients tomography (CT scan) of the brain were recorded.

Patients of 18 years of age or above and hospitalized in our hyperbaric center were prospectively enrolled in the Follow-up study if they met all eligibility criteria. Inclusion criteria included: (1) clinical condition at risk for iatrogenic gas During the hospital stay, data were collected for vital embolism; (2) at least one of the following clinical sign: signs, results from laboratory tests and any major interventions performed. The neurological status, as gas embolism. Among these patients, gas embolism was established by a senior neurologist, and mortality were proven in only 125 cases that were referred to the recorded at discharge from the ICU, at discharge from the hyperbaric unit for HBO (Table 1). Six patients (home- hospital, at 6 months and at 1 year. All patients had visual less) were lost to follow-up after hospital discharge. ﬁeld evaluation tested by Goldman kinetic perimetry and Almost all accidents were air embolism except nine cases brain MRI or CT scan at 6 months and at 1 year following of carbon dioxide gas embolism following coelioscopy or gas embolism. hysteroscopy (Table 2). The main clinical manifestations did not differ between venous and arterial gas embolism, except a higher rate of acute myocardial infarction in case Endpoints of arterial gas embolism (26 vs. 10%, P = 0.03; Table 3).

The primary outcome was 1-year mortality. Secondary outcomes were neurological sequels similarly assessed at Mortality ICU discharge, then among ICU survivors at 6 months and 1 year using the Glasgow Outcome Score (1, death; 2, The crude mortality was 14/119 (12%) at ICU discharge, vegetative state; 3, severe disability, i.e., the patient is 19/119 (16%) at hospital discharge, 21/119 (17.6%) at unable to live independently; 4, moderate disability, i.e. 6 months and 25/119 (21%) at 1-year (Fig. 1). the patient is able to live independently but unable to Univariate analysis suggested that, immediately fol- return to work; 5, mild or no disability, i.e., the patient is lowing gas embolism onset, non-survivors were more able to return to work) [16]. likely to present with cardiac arrest (P = 0.0004), car- diovascular collapse (P = 0.005), acute myocardial infarction (P = 0.02), acute kidney failure (P = 0.01) Statistical analysis and disseminated intravascular coagulopathy (P = 0.045) (Table 4). At the time of ICU admission, non-survivors Data are expressed as median and interquartile range for had lower Glasgow Coma Scale scores (P \ 0.0001), continuous variables and as number and percentage for higher SAPS II scores (P \ 0.0001) and arterial lactate binary variables. Comparisons were made using Student’s levels (P = 0.001), and were more likely receiving t or Mann–Whitney tests for continuous variables (as vasopressor therapy (P = 0.03) and on mechanical appropriate), and chi-square of Fischer’s exact test (as appropriate for binary variables). The occurrences of death and sequels were analyzed independently, based on the same analysis plan, as Table 1 Population characteristics described below. Variables N = 119 Median They were considered as binary outcomes, for which (quartiles), the association with patient, disease and other character- N (percentage) istics was assessed through estimated odds ratio (OR) and 95 percent confidence intervals (95% CI) computed from Age (years) 53 (33–65) Gender (male) 67 (56) fitted logistic regression models. Multivariate logistic Type of gas embolism models were fitted, with goodness of fit assessed by the Venous 92 (77) Hosmer-Lemeshow test, where small P values indicate a Arterial 27 (23) lack of fit of the model [17]. All tests were two-sided, Nature of gas with P values of 0.05 or less denoting statistical Air 110 (92) CO2 9 (8) significance. Known patent foramen ovale 15 (13) Statistical analyses were performed on SAS 9.1 (SAS Brain imaging prior HBO: yes 51 (43) Inc, Cary, NC) and R 2.7.2 (http://www.R-project.org) Time from accident to HBO (h) 6 (5–10) Admission severity of illness software packages. SAPS II 33 (21–55) Glasgow Coma Score 8 (3–15) Mechanical ventilation at admission 76 (64) Vasopressor therapy 27 (23) Outcome Results Duration of coma (h) 10 (0–48) Duration of coma in survivors (h) 30 (12–120) Study population Duration of mechanical 1 (0–5) ventilation (days) From January 1993 to August 2004, a total of 4,727,496 Duration of mechanical ventilation 3 (1–8) in survivors (days) patients were admitted at the Assistance Publique ICU length of stay (days) 3 (2–10) Hoˆpitaux de Paris, and 1,931 (0.04%) had suspicion of Table 2 Causes of gas embolism analysis suggested that cardiac arrest at time of gas embolism and an ICU admission SAPS II score of 33 Clinical context N (%) or more were the only independent prognostic fac- Central venous catheterization 29 (24.3%) tors with an odds ratio of dying of 4.39 (95% CI 1.46– Removal 9 (7.5%) 12.20) and 6.30 (95% CI 1.71–23.21), respectively Accidental removal 8 (6.7%) (Table 5). Manipulation and usual care 7 (5.9%) Among ICU survivors, 1-year non-survivors were Insertion 5 (4.2%) Peripheral venous catheterization 2 (1.7%) older (P = 0.01), were more likely to present with Ba- Thoracic procedures 25 (21%) binski sign, had a longer duration of coma (P = 0.006), Pleural puncture or drainage 8 (6.7%) of mechanical ventilation (P = 0.005) and of ICU stay Trans-thoracic biopsy 4 (3.3%) (P = 0.01) (Table 4). Multivariate analysis suggested as Tumorectomy 4 (3.3%) Bronchoscopy 3 (2.5%) independent prognostic factors age (OR = 1.07, 95% CI Pleuroscopy 2 (1.7%) 1.01–1.14), presence of Babinski sign (OR = 6.58, 95% Miscellaneous 4 (3.4%) CI 1.14–38.20) and presence of acute kidney failure Cardiac surgery 18 (15%) (OR = 8.09, 95% CI 1.28–51.21) (Table 5). Valve surgery 7 (5.9%) Coronary bypass surgery 6 (5%) Inter-auricular communication closure 3 (2.5%) Inter-ventricular communication closure 2 (1.7%) Neurological sequels Abdominal surgery 14 (12%) Laparoscopy 7 (5.9%) Among the 105 ICU survivors, 45 patients (42.9%) left Liver transplant 4 (3.3%) the ICU with neurological sequels, including mainly 4 Miscellaneous 3 (2.5%) Neurosurgery–sitting position 7 (5.8%) (8.9%) patients with vegetative state, 31 (6.7%) with focal Spine surgery 2 (1.7%) motor deficits, 7 (15.6%) with restriction of the visual Facial surgery 1 (0.8%) field, 3 (6.7%) with cognitive problems and 2 (4.4%) with Interventional radiology 13 (10.8) seizures (Table 4). During follow-up, there was a sub- Arteriography 6 (5%) stantial attenuation of sequels at 6 months with fewer Coronarography 4 (3.3%) Miscellaneous 3 (2.5%) patients having a GOS 2, GOS 3 and a GOS 4, and more During renal replacement therapy 5 (4.2%) patients who recovered and resumed to their previous Gynecologic celioscopy 2 (1.7%) social activity (GOS 5) (Fig. 1). Massive blood transfusion 3 (2.5%) Univariate analysis suggested that patients with Suicide attempt (air auto-infusion) 2 (1.7%) Miscellaneous 3 (2.5%) neurological sequels were older (P = 0.01), had a longer delay from accident to HBO therapy (P = 0.03) and were more likely to receive HBO more than 7 h from ventilation (P = 0.02). Non-survivors had longer duration gas embolism (P = 0.02), were more likely to present at of coma (P \ 0.0001), of mechanical ventilation ICU admission with focal motor deficits (P \ 0.0001), (P \ 0.0001) and of ICU stay (P = 0.02). Multivariate Babinski sign (P \ 0.0001), cranial nerve paralysis

Table 3 Main clinical presentation in venous and Symptoms All patients Venous gas Arterial gas P value arterial gas embolism (N = 119) embolism embolism (N = 92) (N = 27)

Neurological symptoms Motor deficit (%) 54 (45%) 44 (48%) 10 (37%) 0.32 Seizure (%) 36 (30%) 26 (28%) 10 (37%) 0.38 Babinski sign (%) 46 (39%) 39 (42%) 7 (26%) 0.12 Visual field abnormalities (%) 16 (13%) 13 (14%) 3 (11%) 0.68 Other cranial nerves abnormalities (%) 35 (29%) 27 (29%) 8 (30%) 0.98 Cardiac symptoms Cardiac arrest (%) 21 (18%) 18 (20%) 3 (11%) 0.32 Chest pain (%) 6 (5%) 6 (7%) 0 0.14 Shock (%) 33 (28%) 25 (27%) 8 (30%) 0.53 Acute myocardial infarction (%) 16 (13%) 9 (10%) 7 (26%) 0.03 Respiratory symptoms Dyspnea (%) 50 (42%) 40 (43%) 10 (37%) 0.48 Cardiogenic pulmonary edema (%) 14 (12%) 10 (11%) 4 (15%) 0.57 Acute lung injury/ARDS 7 (6%) 6 (7%) 1 (4%) 0.58 Others symptoms Acute renal failure (%) 15 (13%) 11 (12%) 4 (15%) 0.69 Disseminated intravascular coagulopathy (%) 11 (9%) 9 (10%) 2 (7%) 0.80 sequels had had brain imaging performed prior to HBO therapy (P = 0.002). Multivariate analysis showed that presence at ICU admission of focal motor deficits (P \ 0.0001) and of Babinski sign (P = 0.0007), and duration of mechanical ventilation larger than 5 days (P = 0.0012) were independent predictors of long-term sequels (Table 5).

Adverse events and observance

Fig. 1 Glasgow Outcome Score (GOS) at intensive care unit (ICU) All patients completed a single HBO session followed by discharge, 6 months and 1 year 12 h of normobaric oxygen therapy, except one patient who had two consecutive HBO sessions because of persistent massive neurological dysfunction and CT scan (P = 0.02), and had longer duration of mechanical evidence of remaining bubbles in the cerebral arteries ventilation (P = 0.006) and ICU stay (P = 0.02) after the ﬁrst HBO session. There were no adverse events (Table 4). Of note, more patients with neurological related to HBO, except one patients who experienced

Table 4 1-year outcome of 119 patients with iatrogenic gas embolism

Variable All patients N = 119 ICU survivors only N = 105 1-year survivors

1-year 1-year Non- P value ICU Survivors 1-year Non- P value Without With P value Survivors survivors (N = 94) survivors sequels sequels (N = 94) (N = 25) (N = 11) (N = 60) (N = 45)

Age (years) 52 (31–63) 57 (44–67) 0.17 52 (31–63) 67 (52–79) 0.01 52 (31–61) 57 (45–70) 0.01 Gender (male) 53 (56) 14 (56) 1.00 53 (56) 5 (45) 0.54 34 (57) 24 (53) 0.84 Type of gas embolism (arterial) 25 (27) 2 (8) 0.06 25 (27) 1 (9) 0.29 17 (28) 9 (20) 0.37 Type of gas (air) 87 (73) 23 (19) 1.00 86 (92) 11 (100) 1.00 55 (92) 43 (96) 0.70 Brain imaging prior to HBO 52 (55) 9 (36) 0.48 52 (55) 3 (27) 0.62 24 (40) 31 (69) 0.002 Evidence for air bubbles at brain imaging 14 (29) 7 (78) 0.008 14 (29) 3 (100) 0.03 4 (18) 13 (43) 0.08 Time to HBO (h) 6 (4–10) 7 (5–10) 0.71 6 (4–10) 7 (4–12) 0.83 6 (4–9.5) 8 (5–12) 0.03 Time to HBO \ 7ha – – – – – – 39 (65) 19 (32) 0.02 Symptoms at time of accident Cardiac arrest 10 (11) 11 (44) 0.0004 10 (11) 2 (18) 0.61 6 (10) 5 (11) 1.00 Mean arterial pressure (mmHg) 60 (14–71) 55 (27–70) 0.98 60 (14–71) 65 (45–89) 0.40 54 (8–70) 69 (40–72) 0.13 Shock 20 (21) 13 (52) 0.005 20 (21) 3 (27) 0.70 14 (23) 9 (20) 0.81 Myocardial infarction 9 (10) 7 (29) 0.02 9 (10) 2 (18) 0.32 6 (10) 5 (11) 1.00 Chest pain 6 (5) 0 (0) 0.34 6 (5) 0 (0) 1.00 5 (8) 1 (2) 0.23 Dyspnea 38 (40) 13 (52) 0.37 38 (40) 8 (73) 0.06 28 (47) 18 (40) 0.55 Pulmonary edema 11 (12) 3 (13) 1.00 11 (12) 2 (18) 0.62 10 (17) 3 (7) 0.15 Focal motor deficit 43 (46) 11 (44) 1.00 43 (46) 8 (73) 0.12 15 (25) 36 (80) \0.0001 Seizures 25 (27) 11 (44) 0.14 25 (27) 4 (36) 0.49 15 (25) 14 (31) 0.51 Babinski sign 36 (38) 10 (40) 1.00 36 (38) 8 (73) 0.05 13 (22) 31 (69) \0.0001 Visual symptoms 15 (16) 1 (4) 0.19 15 (16) 0 (0) 0.19 7 (12) 8 (18) 0.19 Other cranial nerves 28 (30) 7 (28) 1.00 28 (30) 3 (27) 1.00 12 (20) 19 (42) 0.02 Symptoms at ICU admission Glasgow coma score 13 (4–15) 3 (3–6) \0.0001 13 (4–15) 4 (3–15) 0.17 14.5 (4–15) 9 (4–15) 0.44 SAPS II 32 (20–43) 60 (47–68) \0.0001 32 (20–30) 47 (22–63) 0.13 32 (20–39) 36 (22–56) 0.11 SAPS II \ 33 50 (53) 3 (12) 0.0002 – – – – – – Shock 16 (17) 11 (44) 0.007 16 (17) 2 (18) 1.00 13 (22) 5 (11) 0.20 DIC 6 (6) 5 (21) 0.045 6 (6) 0 (0) 1.00 4 (7) 2 (4) 0.70 Acute kidney failure 8 (9) 7 (29) 0.01 8 (9) 0 (0) 0.59 3 (5) 5 (11) 0.28 Creatine kinase (9 normal value) 1(1–4) 2 (1–7) 0.09 1 (1–4) 2 (1–6) 0.38 1 (1–3) 1 (1–4) 0.73 Lactates 1.8 (1.2–2.9) 6.2 (2.6–10) 0.001 1.8 (1.2–2.9) 4.2 (2.6–8.3) 0.05 1.3 (1.2–2.6) 2.7 (1.3–3.7) 0.15 Duration of coma (h) 7 (0–24) 96 (16–240) \0.0001 7 (0–24) 96 (16–240) 0.006 8 (0–18) 20 (0–72) 0.08 Duration of coma \20 h – – – – – – 47 (78) 23 (51) 0.007 Duration of mechanical ventilation (days) 1 (0–3) 5 (2–20) \0.0001 1 (0–3) 5 (2–20) 0.005 1 (0–2) 3 (0–8) 0.006 ICU length of stay (days) 3 (2–8) 7 (3–29) 0.01 3 (2–8) 9 (3–32) 0.01 3 (2–5) 7 (2–19) 0.002 a The value of 7 h was used as it corresponded to the median in non-survivors Table 5 Multivariate analyses for death and neurological Odds ratio, P value sequels in 119 patients with 95% confidence iatrogenic gas embolism intervals All patients—crude 1-year mortality 25/119 SAPS II [33 6.30 [1.71; 23.21] 0.006 Cardiac arrest 4.39 [1.46; 12.20] 0.008 Hosmer-Lemeshow goodness-of-fit test, P = 0.91 ICU survivors—crude 1-year mortality 11/105 Age (years) 1.07 [1.01; 1.14] 0.02 Duration of mechanical ventilation (days) 1.06 [0.99; 1.14] 0.07 Babinski sign 6.58 [1.14; 38.20] 0.04 Acute renal failure 8.09 [1.28; 51.21] 0.03 Hosmer-Lemeshow goodness-of-fit test, P = 0.28 ICU survivors—crude rate of sequels at 1 year 45/105 Duration of mechanical ventilation [5 days 15.14 [2.92; 78.52] 0.0012 Focal motor deficit 12.78 [3.98; 41.09] \0.0001 Babinski sign 6.76 [2.24; 20.33] 0.0007 Hosmer-Lemeshow goodness-of-fit test, P = 0.13 seizures during the HBO session, which resolved by soon as possible, and other treatments may have included shifting the patient from pure oxygen to air. cardiopulmonary resuscitation, volume expansion, mechanical ventilation and positive inotropic support [1, 18, 19]. There is no consensus on the practical modalities for HBO session in iatrogenic gas embolism [20]. Our Discussion treatment procedure consisted of achieving optimal mechanical reduction of bubbles size and optimizing blood This is the first prospective cohort study that included a denitrogenation. Thus, priority was given to the duration of 1-year follow-up of patients with documented iatrogenic oxygen exposure rather than to pressure. In addition, the gas embolism. In this study the crude prevalence of iat- mechanical effect on bubble size reduction may not be rogenic gas embolism was 2.65 cases per 100,000 hospital substantially influenced by increasing the pressure from 4 admissions. Our hyperbaric unit is the only public facility to 6 absolute atmospheres [21–23]. We used high oxygen for the Paris great area, a region of nearly 13 millions pressure (3.8 ATA) and with careful prevention with inhabitants (20% of the French population). The observed benzodiazepine or barbiturates, we did not observe an prevalence of iatrogenic gas embolism is in line with increased risk of seizures. We did not observe any evidence recent reports from the Haute Autorite´ de Sante´ (HAS) in for lung toxicity during the HBO session, and none of the France [3]. This study also established that about one out hospital survivors had pulmonary sequels at the long term. of five patients with iatrogenic gas embolism died within The study population shared most of the characteristics 1 year, and that one out of five survivors had persistent of previously reported cohorts [4–14], highlighting the still neurological sequels at 1 year following the accident. too high proportion of gas embolism following central The design of this prospective inception cohort study venous catheterization [24, 25]. Of note, the clinical pre- differed markedly from previous reports on gas embolism sentation was not significantly different between venous [4–14]. We used strict criteria to define gas embolism and arterial gas embolism except more myocardial ische- including not only the onset of neurological, cardiovas- mia in patients with arterial gas embolism. The observed cular or respiratory symptoms during exposure to a large proportion of patients with neurological symptoms medical or surgical procedure at risk of gas embolism, but following venous gas embolism suggested right-to-left also direct evidence of gas in the systemic circulation shunting via intrapulmonary shunt or patent foramen noticed during or immediately after the procedure. We ovale, found in 15 out of 31 patients who had trans- included only patients with iatrogenic accident and not esophageal echocardiography. This prevalence of patent patients with decompression sickness. All patients were foramen ovale is in line with previous reports [26]. closely followed up for 1 year with a particular focus on Short-term mortality was about 12%, in keeping with neurological sequels as assessed by a senior neurologist mortality rates reported in retrospective studies [4–14]. from clinical examination, visual field tested by Goldman However, after ICU discharge, there were still a signifi- kinetic perimetry and brain imaging. All patients were cant number of deaths with a 1-year mortality of 21%, and treated in a hyperbaric chamber, according to recommen- almost 43% of ICU survivors had neurological sequels. dations from the European Committee for Hyperbaric The source of gas embolism, venous versus arterial, did Medicine or Undersea and Hyperbaric Medical Society not influence the risk of death or long-term neurological [2]. Hyperbaric oxygen therapy should be delivered as sequels, in contrast to previous observations [14]. Similarly, in contrast to common thoughts, the type of In conclusion, iatrogenic gas embolism complicates gas, air versus carbon dioxide, did not influence long-term 2.65 per 100,000 hospitalizations, and is associated with mortality and morbidity. Patients with major arterial gas high long-term mortality and morbidity. The type and the embolism during cardiac or vascular surgery may have source of gas embolism have little impact on outcome, died prior to being referred to the hyperbaric unit. This whereas a time to hyperbaric oxygen therapy of 7 h or study confirmed that the time to HBO treatment is a less reduced the risk of mortality and neurological prognostic factor [14], and in practice a delay of 7 h or sequels. Presence of Babinski signs at time of ICU less was associated with a better outcome. Of note, brain admission is associated with poor long-term outcome. imaging was performed in 43% of patients prior to referral to the hyperbaric unit causing on average a 2 h delay to HBO treatment. Then, unsurprisingly, these patients were more likely to present with long-term neu- Appendix rological sequels. These findings highlighted that when gas embolism is suspected, imaging should not be per- See Fig. 2. formed prior to HBO therapy. Cardiac arrest at the time of accident was the main risk factor for short-term mortality, probably highlighting the Design of an Hyperbaric Oxygen Therapy Session large volume of gas that entered in the circulation. Among 4 risk factors for long-term mortality and sequels, the pres- 3 ence of Babinski signs on ICU admission is probably the most interesting factor. Indeed, examination of toe reflexes 2 is a very easy and robust test. It may allow rapid recog- 1 nition of severe cases and prompt referral to hyperbaric 0 Pressure (ATA) units. To determine whether exposure to high oxygen 0 15 30 45 60 75 90 105 120 135 150 168 pressure had contributed to neurological sequels would Time (min) require comparing our HBO strategy to a strategy of lower Fig. 2 Design of chamber session, showing the pressure and depth and/or lower duration of exposure to oxygen. duration of oxygen exposure

References

1. Muth CM, Shank ES (2000) Gas 7. Murphy B (1985) Cerebral air 12. Boussuges A, Blanc P, Molenat F, embolism. N Engl J Med 342:476–482 embolism resulting from invasive Barthelemy A, Bergmann E, Sainty JM 2. 7th European Consensus Conference on medical procedures. Treatment with (1995) Cerebral air embolism and Hyperbaric Medicine, Lille, France. hyperbaric oxygen. Ann Surg 201:242– central venous catheter. Reán Urg Available via http://www.echm.org/ 245 4:253–259 ECHM-Conferences.htm. Accessed 8. Takahashi H, Kobayashi S, Hayase H, 13. Mathieu D, Wattel F, Embolies Dec 2004 Sakakibara K (1987) Iatrogenic air gazeuses (2002) In: Mathieu D, Wattel 3. Haute Autorite´ de Sante´. Service embolism: a review of 34 cases. F (eds) Traite´ de me´decine hyperbare, e´valuation des actes professionnels. Underwater and hyperbaric physiology 2nd edn. Ellipses, Paris, France, Oxygeńothe´rapie hyperbare. Available IX. In: Proceedings of the ninth pp 202–206 via http://www.has-sante.fr/portail/ international symposium on underwater 14. Blanc P, Boussuges A, Henriette K, jcms/c_498529/oxygenotherapie- and hyperbaric physiology, Bethesda, Sainty JM, Deleflie M (2002) Iatrogenic hyperbare. Accessed 31 Jan 2007 MD. UHMS, pp 931–948 cerebral air embolism: importance of an 4. Goulon M, Barois A, Nouaillhat F, 9. Knauer V, Kopferschmitt J, Mantz JM early hyperbaric oxygenation. Intensive Babinet P, Grosbuis S, Romion A (1989) Facteurs pronostiques des Care Med 28:559–563 (1973) Air embolisms. Rev Prat embolies gazeuses iatroge`nes traiteés 15. Legall JR, Lemeshow S, Saulnier F 23:1541–1560 par l’oxyge`ne hyperbare. A propos (1993) A new Simplified Acute 5. Hart GB (1974) Treatment of d’une se´rie de 27 cas. Bull Med Sub Physiology Score (SAPS II) based on a decompression illness and air embolism Hyp 8:171–177 European/North American multicenter with hyperbaric oxygen. Aerosp Med 10. Massey EW, Moon RE, Shelton D, study. JAMA 270:2957–2963 45:1190–1193 Camporesi EM (1990) Hyperbaric 16. Wilson JTL, Pettigrew LEI, Teasdale G 6. Saulnier F, Durocher A, Dubois D, oxygen therapy of iatrogenic air (1998) Structured interviews for the Fourrier F, Mathieu D, Chopin C, embolism. J Hyperbaric Med 5:15–21 Glasgow Outcome Scale and the Wattel F (1983) Gas embolisms 11. Boussuges A, Blanc P, Molenat F, extended Glasgow Outcome Scale: (apropos of 44 cases). LARC Med Bergmann E, Sainty JM (1995) guidelines for their use. J Neurotrauma 3:131–134 Prognosis in iatrogenic gas embolism. 15:573–585 Minerva Med 86:453–457 17. Hosmer DW, Lemeshow S (1989) 20. Fukaya E, Hopf HW (2007) HBO and 24. Ely EW, Hite RD, Baker AM, Johnson Assessing the fit of the model. In: gas embolism. Neurol Res 29:145–149 MM, Bowton DL, Haponik EF (1999) Hosmer DW, Lemeshow S (eds) 21. Leitch D, Greenbaum J, Hallenbeck J Venous air embolism from central Applied logistic regression. Wiley, (1984) Cerebral arterial air embolism: venous catheterization: a need for New York, pp 135–175 III. Cerebral blood flow after increased physician awareness. Crit 18. Alvaran SB, Toung JK, Graff TE, decompression from various pressure Care Med 27:2113–2117 Benson DW (1978) Venous air treatments. Undersea Biomed Res 25. McGee DC, Gould MK (2003) embolism: comparative merits of 11:249–263 Preventing complications of central external cardiac massage, intracardiac 22. Leitch D, Greenbaum J, Hallenbeck J venous catheterization. N Engl J Med aspiration, and left lateral decubitus (1984) Cerebral arterial air embolism: I. 348:1123–1133 position. Anesth Analg 57:166–170 Is there benefit in beginning HBO 26. Hagen P, Scholz D, Edwards W (1984) 19. Annane D, Troche´ G, Delisle F, treatment at 6 bars? Undersea Biomed Incidence and size of patent foramen Devauchelle P, Paraire F, Raphae¨l JC, Res 3:221–235 ovale during the first 10 decades of life: Gajdos P (1994) Effects of mechanical 23. Mc Dermott J, Dutka A, Koller W, an autopsy study of 965 normal hearts. ventilation with normobaric oxygen Pearson R, Flynn E (1992) Comparison Mayo Clin Proc 59:17–20 therapy on the rate of removal from of two recompression profiles in cerebral arteries. Crit Care Med treating experimental cerebral air 22:851–857 embolism. Undersea Biomed Res 19:171–185